Robust Estimation of Landslide Displacement from Multi-temporal UAV Photogrammetry-Derived Point Clouds

• Published in: IEEE journal of selected topics in applied earth observations and remote sensing 2024-03, p.1-16
• Main Authors: He, Haiqing, Ming, Zaiyang, Zhang, Jianqiang, Wang, Leyang, Yang, Ronghao, Chen, Ting, Zhou, Fuyang
• Format: Article
• Language: English
• Subjects: Autonomous aerial vehicles; error propagation; Estimation; intra- and inter-estimates; Landslide displacement; Point cloud compression; robust estimation; Rough surfaces; Surface morphology; Surface roughness; Terrain factors; unmanned aerial vehicle
• ISSN: 1939-1404
• DOI: 10.1109/JSTARS.2024.3373505

Existing algorithms based on remote sensing for landslide displacement estimation, such as C2C, C2M, DOD, and M3C2, are sensitive to errors generated in data processing, and further improving their accuracy is difficult. To address this issue, given that redundant observations may occur in landslide monitoring, we proposed a robust estimation method of landslide displacement from multi-temporal unmanned aerial vehicle (UAV) photogrammetry-derived point clouds. The proposed method first establishes the relevant graph to manage the trajectory of error propagation for landslide displacement estimation for all possible paths. Two modules, namely, intra- and inter-estimates, are explored to reduce the impact of outliers and high surface roughness in point clouds derived by UAV photogrammetry. Individually, the intra-estimate operation is used to calculate landslide displacement between two temporal point clouds by robust estimation considering outlier constraint, and the inter-estimate operation is used to obtain the optimal calculation of landslide displacement by minimizing a given objective function using IGG robust estimation proposed by the Institute of Geodesy and Geophysics at the Chinese Academy of Sciences. Experimental results show that the proposed method is significantly superior to conventional methods such as C2C, C2M, and M3C2, with an accuracy improvement of at least 8%.